Enum AgentEvent 

pub enum AgentEvent {
    UserTaskSubmitted {
        text: String,
        content_parts: Vec<ContentPart>,
        submitted_at: Option<u64>,
    },
    ModelOutput {
        text: String,
        thinking: Option<String>,
    },
    ModelDelta {
        text: String,
    },
    ModelThinkingDelta {
        text: String,
    },
    ToolRequested(ToolInvocation),
    ToolCompleted(ToolResult),
    SubagentActivity {
        invocation_id: String,
        message: String,
    },
    SubagentTranscript {
        invocation_id: String,
        item: SubagentTranscriptItem,
    },
    HarnessTrace(Value),
    HarnessStopped {
        reason: String,
        message: String,
        tool_name: Option<String>,
    },
    PatchProposed(PatchProposal),
    ApprovalRequested(ApprovalRequest),
    ApprovalResolved(ApprovalDecision),
    CapabilityRecorded(CapabilityLedgerEntry),
    QuestionRequested(QuestionRequest),
    QuestionResolved(QuestionResponse),
    PlanReviewRequested(PlanReviewRequest),
    PlanReviewResolved(PlanReviewResponse),
    SudoPasswordRequested(SudoPasswordRequest),
    RepeatedToolCallWarning {
        tool_name: String,
        message: String,
    },
    RepetitionDetected {
        kind: RepetitionWarningKind,
        message: String,
    },
    Error {
        message: String,
    },
    AutoDreamStarted {
        hours_since_last: u64,
        sessions_reviewed: usize,
    },
    AutoDreamCompleted {
        marked_stale: usize,
        duplicates_merged: usize,
        active_count: usize,
    },
    AutoDreamFailed {
        reason: String,
    },
    UsageReported {
        input_tokens: u64,
        output_tokens: u64,
        cache_creation_tokens: u64,
        cache_read_tokens: u64,
    },
    SessionRecap {
        summary: String,
        suppressed: bool,
    },
    StreamResuming {
        accumulated_chars: usize,
        attempt: u32,
    },
    SetGoalRequested {
        objective: String,
        short_description: Option<String>,
        token_budget: Option<i64>,
    },
    GoalUpdated {
        session_id: String,
        goal_id: String,
        objective: String,
        short_description: Option<String>,
        status: GoalStatus,
        tokens_used: i64,
        token_budget: Option<i64>,
    },
    MicroCompactApplied {
        messages_cleared: usize,
    },
    AutoCompactStarted,
    AutoCompactCompleted {
        tokens_saved: u64,
    },
    AutoCompactFailed {
        reason: String,
    },
    PlanProposed {
        title: String,
        steps: Vec<String>,
    },
    AgentModeChanged {
        mode: AgentMode,
    },
    NotificationRequested {
        title: String,
        body: String,
        urgency: NotificationUrgency,
        category: Option<String>,
    },
    UpdateAvailable {
        current_version: String,
        latest_version: String,
        latest_tag: String,
        release_url: String,
        body: Option<String>,
        prerelease: bool,
    },
}

A high-level agent event suitable for client consumption.

Unlike RuntimeEventKind, agent events represent the semantic actions a client cares about: user input, model output, tool calls, approvals, compaction, usage, and errors.

Variants

UserTaskSubmitted

The user submitted a new task or message.

Fields

text: String

The user’s input text.

content_parts: Vec<ContentPart>

Optional multimodal content parts (images + text).

submitted_at: Option<u64>

Unix timestamp (seconds since epoch) when the user submitted this message. Used for wall-clock display in clients (e.g. TUI sticky bar). Optional for backward-compatible session JSON (pre-timestamp sessions).

ModelOutput

A complete model output with optional thinking/reasoning content.

Fields

text: String

The assistant’s response text.

thinking: Option<String>

Optional thinking or reasoning trace from the model.

ModelDelta

A streaming text delta from the model.

Fields

text: String

Incremental text content.

ModelThinkingDelta

A streaming thinking/reasoning delta from the model.

Fields

text: String

Incremental thinking content.

ToolRequested(ToolInvocation)

The assistant requested a tool invocation.

ToolCompleted(ToolResult)

A tool invocation completed.

SubagentActivity

Transient status for a nested subagent.

Fields

invocation_id: String

Parent subagent tool invocation id.

message: String

Human-readable description of the latest nested activity.

SubagentTranscript

Transient drill-down transcript item for a nested subagent.

Fields

invocation_id: String

Parent subagent tool invocation id.

item: SubagentTranscriptItem

Transcript item to append for this UI session.

HarnessTrace(Value)

A harness-level diagnostic trace.

HarnessStopped

The harness stopped a turn before another model iteration.

Fields

reason: String

Machine-readable stop reason.

message: String

Human-readable diagnostic.

tool_name: Option<String>

Tool involved in the stop, when applicable.

PatchProposed(PatchProposal)

A file patch was proposed by the assistant.

ApprovalRequested(ApprovalRequest)

A tool invocation requires user approval.

ApprovalResolved(ApprovalDecision)

An approval request was resolved.

CapabilityRecorded(CapabilityLedgerEntry)

A capability lifecycle event was recorded by the policy layer.

QuestionRequested(QuestionRequest)

The assistant requested an interactive user choice.

QuestionResolved(QuestionResponse)

An interactive user choice was resolved.

PlanReviewRequested(PlanReviewRequest)

Plan tool create is waiting for user review (blocks the turn).

PlanReviewResolved(PlanReviewResponse)

User finished plan review (approve / changes / quit).

SudoPasswordRequested(SudoPasswordRequest)

Bash/sudo needs a password; TUI shows a masked modal. Password is never included in this event — only a correlation id. The secret is delivered solely through the sudo password resolver oneshot.

RepeatedToolCallWarning

The same tool was called consecutively with identical arguments. The tool is NOT executed; this is a notification to the user.

Fields

tool_name: String

Name of the repeated tool.

message: String

Warning message describing the repetition.

RepetitionDetected

Repetitive/degenerate model output was detected (character runs, alternating patterns, or duplicate thinking blocks).

Fields

kind: RepetitionWarningKind

What kind of repetition was detected.

message: String

Human-readable warning message.

Error

An error occurred.

Fields

message: String

Human-readable error message.

AutoDreamStarted

Auto-dream memory consolidation has started.

Fields

hours_since_last: u64

Hours since the last dream run.

sessions_reviewed: usize

Number of sessions reviewed.

AutoDreamCompleted

Auto-dream memory consolidation has completed.

Fields

marked_stale: usize

Memories marked stale.

duplicates_merged: usize

Duplicates merged.

active_count: usize

Active memories remaining.

AutoDreamFailed

Auto-dream memory consolidation has failed.

Fields

reason: String

Human-readable reason for the failure.

UsageReported

Token usage was reported by the model provider.

Fields

input_tokens: u64

Number of input/prompt tokens consumed.

output_tokens: u64

Number of output/completion tokens produced.

cache_creation_tokens: u64

Number of tokens written to the prompt cache (Anthropic).

cache_read_tokens: u64

Number of tokens read from the prompt cache (Anthropic).

SessionRecap

Short post-turn session recap (“Recap” line).

Fields

summary: String

One- or two-sentence summary of the turn.

suppressed: bool

When true, the recap was generated but should not be shown (long-tail).

StreamResuming

The provider stream broke mid-generation and is being resumed via prefill (assistant continuation). The UI can show a transient hint.

Fields

accumulated_chars: usize

Characters of text accumulated before the break.

attempt: u32

Retry attempt number (1-based).

SetGoalRequested

The agent requested to set a goal via natural language.

Fields

objective: String

The objective text.

short_description: Option<String>

Optional short UI label.

token_budget: Option<i64>

Optional token budget.

GoalUpdated

The session goal was updated (created, status change, budget exceeded).

Fields

session_id: String

The session this goal belongs to.

goal_id: String

Unique identifier for the goal.

objective: String

The objective text.

short_description: Option<String>

Optional short UI label.

status: GoalStatus

Current goal status.

tokens_used: i64

Tokens consumed so far.

token_budget: Option<i64>

Optional token budget.

MicroCompactApplied

Micro-compaction cleared stale tool results from history.

Fields

messages_cleared: usize

Number of tool result messages cleared.

AutoCompactStarted

Automatic conversation compaction started.

AutoCompactCompleted

Automatic conversation compaction completed.

Fields

tokens_saved: u64

Estimated tokens saved by compaction.

AutoCompactFailed

Automatic conversation compaction failed.

Fields

reason: String

Human-readable failure reason.

PlanProposed

The agent proposed a plan in Plan mode.

Fields

title: String

Title/summary of the plan.

steps: Vec<String>

Ordered list of steps.

AgentModeChanged

The agent mode changed (e.g. Default → Plan or Plan → Default).

Fields

mode: AgentMode

The new mode.

NotificationRequested

Host should surface a user-visible notification.

Desktop hosts call OS toasts; browser hosts map this to the Web Notifications API. Payload mirrors crate::notify::NotifyRequest.

Fields

title: String

body: String

urgency: NotificationUrgency

category: Option<String>

UpdateAvailable

A newer NAVI release is available (from update check).

Fields

current_version: String

latest_version: String

latest_tag: String

release_url: String

body: Option<String>

prerelease: bool

Trait Implementations

impl Clone for AgentEvent

fn clone(&self) -> AgentEvent

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for AgentEvent

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl<'de> Deserialize<'de> for AgentEvent

fn deserialize<__D>(__deserializer: __D) -> Result<Self, __D::Error>

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl Serialize for AgentEvent

fn serialize<__S>(&self, __serializer: __S) -> Result<__S::Ok, __S::Error>

where __S: Serializer,

Serialize this value into the given Serde serializer.